How to self-host Flowise in 2026: your chatbot, your server, your API keys

What Flowise is and why people self-host it

Flowise is a visual builder for AI: drag nodes onto a canvas — a model, a data source, a memory, a tool — wire them together, and you've built a chatbot or an agent without writing framework code. Every flow you build exposes two things: an embeddable chat widget you can paste into any site, and an API endpoint you can call from an app. 55,000+ GitHub stars, Apache-2.0 licence, moving fast.

Skip the manual setup: livemy.app has a one-click Flowise template — free SSL, persistent storage, and a live canvas in about two minutes.

Three reasons to run it yourself rather than pay a chatbot platform:

• No per-bot, per-message fees. Chatbot builders charge monthly per bot and meter your messages. Self-hosted Flowise runs unlimited flows for the price of the server, and predictions are billed at cost by your model provider — not marked up by a platform in between.

• Your data trains your bot. The documents you feed it — support docs, product manuals, internal wikis — and every visitor conversation stay in your own database.

• One instance, many bots. A support bot, a lead-capture assistant and an internal Q&A tool are three flows on the same server, not three subscriptions.

This is the same economics that makes self-hosted LibreChat work — the interface is free and open source, and the AI itself is billed at cost. Flowise is what you reach for when the goal is a bot for your visitors rather than a chat tool for your team.

Two ways to self-host Flowise

Path A: Docker on your own server

Flowise ships Docker images and a Compose setup. The short version: pull the image, set your environment variables, start the container, open the port. What the quick-start leaves out is the production wrapper — a reverse proxy with SSL, persistent volumes so your flows survive a rebuild, authentication before the URL is public, and updates (Flowise releases often). Budget 1–2 hours for a proper first deploy.

Path B: livemy.app one-click template

Pick the Flowise template, deploy, and the canvas is live with SSL handled. Add your model API keys under Project Settings → Environment Variables, build your first flow, paste the embed snippet into your site. Flat $10/month on the Maker plan with a custom domain; backups are a $5/month add-on — worth it once real flows and documents live in there.

Five things that quietly break self-hosted Flowise

1. The canvas is public and unauthenticated

A fresh Flowise on a public URL is an open door: anyone who finds it can open your flows, read how your bot works, and — worse — run predictions that bill your API keys.

Fix. Turn on authentication before the URL leaves your browser history. Set the app-level login credentials in your environment variables, restart, and confirm the canvas asks for a password from an incognito window.

2. Flows vanish after a redeploy

Flowise stores flows, credentials and settings in its database. If that storage isn't on a persistent volume, the next container rebuild resets the instance to zero — and a canvas full of carefully wired flows is gone.

Fix. Verify persistence before building anything real: create a test flow, redeploy, check it's still there. On the livemy.app template persistent storage is wired up; on DIY Docker, mount the data directory as a volume.

3. No spending caps on the API keys

A public-facing chatbot spends your model budget on every visitor message. A pricing bug in your flow, a bot that loops, or plain popularity can turn into an unpleasant invoice from your model provider.

Fix. Set hard spending limits in the provider's dashboard (OpenAI, Anthropic, Google all support them). Use a cheap model for high-volume public flows and save the expensive one for flows where quality is the point.

4. The document index disappears on restart

“Chat with your docs” flows store embeddings in a vector store. The default in-memory option is fine for testing and silently empty after every restart — the bot suddenly answers as if it never read your documents.

Fix. For production, use a persistent vector store node instead of in-memory, and re-index after major changes. If the bot's answers get vague after a redeploy, this is the first thing to check.

5. The embed widget left open to any domain

The chat widget calls your instance from the visitor's browser. Left unrestricted, anyone can embed your bot — and your API budget — on their site.

Fix. Restrict which domains may embed each flow in the flow's configuration, and keep the API endpoint behind its key for server-side calls.

What it costs

Chatbot platforms typically price per bot per month, with message caps that push you up tiers as the bot gets used. Self-hosted Flowise inverts that:

• Hosting — $10/month flat on livemy.app Maker (custom domain and SSL included), or $5–15/month on a VPS you manage yourself.

• Predictions — at cost. Billed by your model provider per token, with caps you control. A modest support bot usually runs a few dollars a month; heavy traffic scales linearly and visibly.

• Bots — unlimited. Additional flows cost nothing. This is where the platform-fee comparison falls apart fastest.

FAQ

Is Flowise free?

The software is free and open source (Apache-2.0). Self-hosted, your costs are the server (~$10/month) and the model API usage your flows generate, billed at cost by the provider. There are no per-bot or per-message platform fees.

Do I need to know LangChain or how to code?

No — the canvas replaces the code. You connect blocks: a model, a data source, a memory, a tool. Paste your model API key and the flow runs. Code is optional, not required.

Can I embed a Flowise chatbot on my website?

Yes — every flow gets an embeddable chat widget (a small script you paste into your site) and an API endpoint, both served from your instance. Restrict the allowed domains before going live.

What does Flowise need to run?

Docker, persistent storage for its database and files, and roughly 2 GB of RAM for typical use. Plus at least one model provider API key — Flowise is the builder, not the model.

How is Flowise different from LibreChat or Dify?

LibreChat is a chat interface for your team to use models. Flowise builds bots and agents for your users, visually. Dify covers similar ground to Flowise with more of a product-platform feel and a steeper footprint. All three are in the livemy.app template gallery, so trying them side by side costs nothing but an afternoon.

Do visitor conversations stay private?

Conversation history and your documents live in your own database. Message content still goes to the model provider you connected, under that provider's API terms — connect a local model if nothing may leave your server.

Will it auto-update?

On livemy.app, redeploy is one click in the dashboard — or trigger it over MCP. Flowise releases frequently; read the notes before major version jumps, and back up first.

The bot is free — budget for the two minutes of security

Flowise collapses the distance between “I want a chatbot on my site” and having one from weeks to an afternoon, and self-hosting it removes the platform fee that usually comes with that convenience. The only real tax is the security pass: authentication on, spending caps set, embed domains restricted. Do those three before sharing the URL and the rest is drag, drop, ship.

→ Deploy Flowise on livemy.app · One-click template · Flat $10/month · Custom domain and SSL included.